Sharma et al. (2026) A multivariate framework for uncertainty quantification in climate-driven streamflow and flood modelling for Tehri Dam catchment of the Indian Himalayas

Identification

• Journal: Stochastic Environmental Research and Risk Assessment
• Year: 2026
• Date: 2026-01-01
• Authors: Bhanu Sharma, N. K. Goel
• DOI: 10.1007/s00477-025-03161-3

Research Groups

• Department of Hydrology, IIT Roorkee, Roorkee, India
• International Centre of Excellence for Dams, IIT Roorkee, Roorkee, India

Short Summary

This study develops a comprehensive multivariate framework for streamflow projections, uncertainty quantification, and flood risk assessment in the Tehri Dam catchment of the Indian Himalayas. It reveals that interactions between climate models (GCMs) and Shared Socioeconomic Pathways (SSPs) contribute over 50% of total streamflow variance, with extreme flood magnitudes projected to nearly double under high-emission scenarios.

Objective

• To develop a comprehensive multivariate framework for streamflow projections, uncertainty quantification, and flood risk assessment in the Tehri Dam catchment, Indian Himalayas.
• To explicitly incorporate interactive uncertainty effects from hydrological model structure, climate models, and socioeconomic scenarios using Multivariate Analysis of Variance (MANOVA) and Principal Component Analysis (PCA).
• To estimate future flood magnitudes using the Gumbel distribution and generate floodplain maps through 2D HEC-RAS modelling to support climate-resilient water resources management.

Study Configuration

• Spatial Scale: Tehri Dam catchment (7293 km²), including 2042 km² of snow-covered peaks and glaciers. Floodplain mapping focused on a 45 km stretch of the Bhagirathi River from Tehri Dam to Devprayag. Meteorological data at 0.25° × 0.25° (rainfall) and 1° × 1° (temperature) for IMD, and 0.25° × 0.25° for CMIP6 bias-corrected data. Digital Elevation Model (DEM) at 30 m resolution, resampled to 1 m for hydraulic modelling.
• Temporal Scale:
  • Historical meteorological data: Daily (IMD).
  • Climate projections: Daily (CMIP6, historical 1951–2005, future 2006–2100).
  • Hydrological model calibration: 2006–2016.
  • Hydrological model validation: 2017–2020.
  • Future streamflow projections: 2021–2100.
  • Extreme value analysis: Observed (2006–2020), Projected (2021–2100).

Methodology and Data

• Models used:
  • Hydrological: Semi-distributed hybrid conceptual hydrological model (GR4J coupled with a snow–glacier module).
  • Uncertainty Quantification: Multivariate Analysis of Variance (MANOVA), Principal Component Analysis (PCA).
  • Extreme Value Analysis: Gumbel distribution (EV-I).
  • Hydraulic Modelling: 2D HEC-RAS.
• Data sources:
  • Meteorological Forcing (Historical): India Meteorological Department (IMD) gridded daily precipitation (0.25° × 0.25°) and daily maximum/minimum temperature (1° × 1°).
  • Future Climate Forcing: Bias-corrected and downscaled CMIP6 data (13 Global Climate Models) for South Asia (0.25° × 0.25° daily) under four Shared Socioeconomic Pathways (SSP126, SSP245, SSP370, SSP585).
  • Topography: Advanced Land Observing Satellite (ALOS) Digital Elevation Model (DEM) (30 m spatial resolution), resampled to 1 m for HEC-RAS.
  • Observed Streamflow: Daily data at Tehri Dam for model calibration and validation.

Main Results

• The hybrid GR4J–snow–glacier hydrological model achieved satisfactory performance with Nash–Sutcliffe Efficiency (NSE) values of 0.766 (calibration) and 0.719 (validation), and R² values of 0.743 (calibration) and 0.742 (validation).
• Multivariate Analysis of Variance (MANOVA) revealed that climate model (GCM)–SSP interactions are the most significant contributor to uncertainty, accounting for 50–60% of total streamflow variance. Three-way interactions (Hydrological model–GCM–SSP) contribute 25–30%.
• Principal Component Analysis (PCA) indicated that the first two principal components explain over 70% of the total variance, with hydrological response (PC1, 42%) and precipitation variability (PC2, 30%) identified as dominant factors.
• Extreme value analysis using the Gumbel distribution projected substantial increases in flood magnitudes; the 1000-year flood discharge is estimated to reach 11,728 m³/s under SSP585, nearly double the observed 5782 m³/s.
• 2D HEC-RAS hydraulic simulations demonstrated significant expansions of inundation zones and elevated water surface elevations (ranging from 440 m to 642.89 m) under high-emission scenarios, identifying critical risk zones such as Maletha and Chaka villages.

Contributions

• Develops a novel, comprehensive multivariate framework that explicitly quantifies and incorporates the interactive effects of multiple uncertainty sources (hydrological model, GCMs, SSPs) in climate-driven streamflow and flood modelling, addressing a critical gap in previous univariate approaches.
• Provides a robust, end-to-end methodology for climate-resilient flood risk management in complex, snow-fed Himalayan catchments, integrating a hybrid hydrological model, advanced uncertainty analysis (MANOVA, PCA), extreme value theory, and 2D hydraulic modelling.
• Offers a scientifically rigorous decision-support tool for water resources management, reservoir operation, and disaster preparedness in mountainous regions highly vulnerable to climate change impacts.
• Highlights the dominant role of GCM-SSP interactions and hydrological response/precipitation variability in overall streamflow projection uncertainty, guiding future research and adaptation strategies.

Funding

• THDC India Limited (THDCIL) for providing data.

Citation

@article{Sharma2026multivariate,
  author = {Sharma, Bhanu and Goel, N. K.},
  title = {A multivariate framework for uncertainty quantification in climate-driven streamflow and flood modelling for Tehri Dam catchment of the Indian Himalayas},
  journal = {Stochastic Environmental Research and Risk Assessment},
  year = {2026},
  doi = {10.1007/s00477-025-03161-3},
  url = {https://doi.org/10.1007/s00477-025-03161-3}
}